There is marked variability in drug response among individuals. This variability poses a major clinical problem by causing decreased drug efficacy or unexpected toxicity. The ability to incorporate predictors of drug efficacy or toxicity into clinicl practice would be a major advance. Well-defined genetic variations in drug metabolizing enzymes or drug targets contribute to variability in drug concentration and therefore response. These pharmacogenetic findings define a predictable component of variability in drug response among individuals. However, despite intensive research and robust findings, there has been almost no translation of pharmacogenetic findings into clinical practice. A critical barrier is tha the importance of pharmacogenetics has not been demonstrated for important patient outcomes in clinical practice. It is not feasible perform a randomized clinical trial to test the clinical importance of every pharmacogenetic question. We propose a novel approach: to use an electronic medical record (EMR) with de-identified information linked to a DNA biobank to test the clinical importance of pharmacogenetic findings for important outcomes of drug therapy in clinical practice. We will implement this novel approach, demonstrating proof-of-principle for three distinct pharmacogenetic findings. These have been chosen for study because there is already overwhelming evidence of an effect on drug metabolism, and consequently pharmacokinetic or pharmacodynamic measures, but genotyping is not yet routine in clinical care. We will test the hypotheses that: 1) CYP2C9 and VKORC1 variants associated with reduced warfarin dose-requirements are associated with greater fluctuation in INR after the warfarin dose-titration phase; 2) CYP2C9 variants with reduced function are associated with more frequent hypoglycemia with sulfonylureas; 3) CYP2D6 poor- and intermediate-metabolizer patients have reduced analgesic effects after receiving codeine for pain. Our approach is to define clinically important pharmacogenetic questions and to test their clinical importance using a combination of bioinformatic, epidemiologic and genetic expertise in a large EMR of more than 1.5 million patients linked to a DNA bank of >157,719 DNA samples. These studies will have high public health impact, not only in translating pharmacogenetic findings to improved patient care for the drugs studied, but also in developing new approaches to testing the importance of future pharmacogenetic observations in clinical practice.